| { |
| "paper_id": "J86-2006", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T02:55:44.545189Z" |
| }, |
| "title": "", |
| "authors": [], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "", |
| "pdf_parse": { |
| "paper_id": "J86-2006", |
| "_pdf_hash": "", |
| "abstract": [], |
| "body_text": [ |
| { |
| "text": "As a programming language for computational linguistics, Prolog is a relative newcomer. Mark Wallace, however, demonstrates very clearly in this timely book the value of this important tool, especially as it relates to the building of natural language front ends and interfaces to database systems. He takes a very practical approach which should appeal to anyone who has had to contend with the difficulties of designing and implementing natural language interfaces.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Wallace begins by providing some background on natural language interfaces. He surveys most of the conceptual issues, but generously intersperses concrete references to major research papers and projects. Some might find his survey too shallow and broad in certain respects, but I personally found his treatment fair and complete.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "In subsequent chapters, Wallace introduces a formal query language, called D&Qs, based on referring phrases (Descriptions) and qualifying phrases (Qualifiers). He then uses this formalism as a representational vehicle in the development and Prolog implementation of a natural language interface, called QPROC. D&Qs is based on predicate calculus, suitably restricted to provide an adequate relational query language. Queries in D&Qs can either be cast into Prolog (as in his \"pilot\" version) or converted by Prolog to an underlying query language. In the Prolog version, each simple qualifier is handled through facts, each relation maps into a predicate, and each tuple of the relation ends up as a Prolog clause for that predicate.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Although the parser is treated in a domain-independent fashion, semantics adopts a fairly conventional relation and attribute style, with verbs, of course, playing the major roles. Some major issues of semantics are clearly identified for the reader, including ambiguity, several matters involving reference and qualification, and ways to handle the verb to be.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The reader should not, however, view this book as something it does not claim to be -namely, a book that provides a thorough and adequate introduction to the areas of database systems, natural language understanding, or Prolog (although the latter is discussed at some length in an appendix). Such readers will be disappointed. A moderate level of competence in these areas is certainly assumed. The interested reader should find the extensive bibliography also very useful. I recommend this book to those who feel competent with Prolog, have a basic understanding of relational database systems, and can reason through a modest amount of predicate calculus. One of the curiosities of practical linguistics has long been the chasm between linguistics, as practiced by, for example, computational linguists, and linguistic theory. Lexical-functional grammar deserves some attention from the practical linguist if only because it contains an explicit attempt to bridge this chasm.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Lexical-functional grammar is firmly planted in Chomskyan tradition.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stan C. Kwasny", |
| "sec_num": null |
| }, |
| { |
| "text": "It builds on all the accepted conclusions of that theory through X-bar theory. Bresnan began lexical-functional grammar in a 1978 publication, and she has continued it since then in a number of publications. The Mental Representation of Grammatical Relations brings together this work and the work of a number of collaborators into a synthesis that includes many interesting innovations. Lexical-Functional Grammar is an individual work that is not \"orthodox\", in the sense that it deviates from Bresnan's own theory in some non-trivial ways. In this review, I will concentrate on Bresnan's theory and discuss Horn's contributions only in passing.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stan C. Kwasny", |
| "sec_num": null |
| }, |
| { |
| "text": "I will begin by trying to describe lexical-functional grammar for the outsider. It will be necessary to squeeze a lot into a few sentences, but the reader needs to understand some of the machinery (for instance, the striking up-arrow and down-arrow notation) to understand anything about the theory.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stan C. Kwasny", |
| "sec_num": null |
| }, |
| { |
| "text": "Grammar is divided into several modules that function independently of one another to process an utterance out of the mind into speech, or in the other direction. Theoreticians like to ignore the production of sentences and move directly to a module, having no generally", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Stan C. Kwasny", |
| "sec_num": null |
| }, |
| { |
| "text": "The Mental Representation of Grammatical l~elations agreed-upon name, which generates all possible constituent structures by exercising a context-free phrase structure grammar (or an equivalent such as a recursive transition network). This leads to a string of lexical categories.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Then a lexical module takes the category string as input and fills each category with a choice from the lexicon. This leads to a string of (abstract) morphemes. Then a phonological module takes the morpheme string and converts it into an utterance. Various schools of thought have inserted transformational modules between the lexical and phonological modules or even between phrase structure and lexicon.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "This model of speech has been criticized for leaving no place for meaning. Chomsky has defended it by insisting that it is a model of grammar and not a model of human thought; there is good reason to believe that he thinks that it is a tool, perhaps the only tool, for examining thought scientifically. But even government-and-binding theory has never really addressed this problem. The generative semanticists identified semantics with the phrase structure module; Chomsky explicitly rejected this idea and, probably, there are, nowadays, no actual generative semanticists left. Lexical-functional grammar is another attempt, quite different in spirit, to address the problem of meaning.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "In lexical-functional grammar, the lexical module is expanded to produce a second output -a functional structure -as well as the constituent structure that it passes to the phonology. The functional structure is the input to a semantic module that determines the meaning. The process of producing an utterance remains shadowy, but it is very clear how understanding must be carried out. In order to drive the lexical functional processing, the phrase structure is enriched with a set of functional statements that parallel the category statements.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "For example, the initial phrase-structure rule of English is expanded into:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "S -~ NP VP (}subj)=} ~--$", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "In this notation, the ~ is a variable to be filled with the identifier of the node being expanded (here S) and the is a variable to be filled with the identifier of the new node (here NP or VP). An interesting effect of this particular rule is the equation S --VP, so that the VP is identified with the sentence and its parts become sentence parts. I presume that VP is retained in the analysis because it reappears elsewhere in the phrase structure, but it clearly has no separate part to play in the meaning.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Hence as an utterance leaves the phrase structure it has a conventional category string and a set of equations like these:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "(flsubj)=f2; fl =f3", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "The lexical module adds more of these equations on the basis of the lexical items it inserts into the category string. For example:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "(f2 pred) = girl", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "The entire set of equations is a functional description, and it must be solved to provide a functional structure. The word \"function\" is not being used lightly here. It is the intent of the theory that all the unknowns, the various fs, are indeed functions in the technical sense of that word. That is, they are sets of ordered pairs of argument and value such that each argument occurs no more than once.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "The description has been solved when each unknown function has been worked out and the value corresponding to each argument determined. In general, this is not too hard if the sentence is well-formed. And a failure in the solution indicates that the sentence is not well-formed. Grammaticality is automatic.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "What makes the functional structure interesting is that a value can be another function. For example, the entire sentence might be a function with four arguments, subj, obj, tense, pred, and the value of the argument subj might be a three-place function with arguments spec, num, pred,", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "where the three values are a, singular, girl. This has advantages; in a reflexive sentence the same function can be the value for both of the arguments subj and obj. In more complex sentences the interrelationship can be quite complicated.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "The claim made by lexical-functional grammar is that functional structure is an adequate representation of mental processes. This is a strong claim that will need a significant amount of testing before it can be accepted. Since a number of persons appear to already believe that thought is, in fact, based on a mixture of predicates and their arguments, this is not an outrageous claim. Lexical-functional grammar can support a clearly formulated version of predicate and argument logic. It can also support considerably more complicated structures unless, as all of our writers are careful to do, one argument of every function is a distinguished \"predicate\". This is not however required by any other part of lexical-functional grammar, and represents a separate claim.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Lexical-functional grammar has no transformations.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "This means that it must face the problems of anaphorical behavior that government and binding was developed to handle. Some of this is relatively easy because relationships that are distant in terms of phonological strings can easily turn up as immediate after the functional description is solved. But neither Bresnan nor Horn can solve all the problems in this way, and they have to fall back onto special devices. The reader need not be concerned about specific devices, because they are sure to be changed in each treatment of the subject. Some transformations may yet be re-introduced. Another difficulty which is almost trivial but matters because it impacts on the name of the theory is that \"function\" is probably a misnomer. One situation Bresnan answers with a special device is the component, like an English temporal adjunct, which can occur an arbitrary number of times. The device used is to allow a value to be a set of functions rather than just one. This is a rather transparent trick to avoid saying that \"adjunct\" occurs several times as an argument and the set of ordered pairs is not, in fact, a function. Either we have to do with something we might call a \"near-function\", or we have to reformulate the well-formedness rules.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Both of the books being reviewed are mainly devoted to proving that this framework does not fall apart when presented with certain standard problems. Horn's book is less rigorous and closer to conventional linguistic theory. The main difference in the two theories seems to be that Horn admits function predicates (that is, values of the argument \"predicate\") that are themselves functions rather than atomic logical predicates. This is interesting from the mental representation point of view and deserves more discussion from that point of view than it gets in either book.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Bresnan has collected a number of very competent co-workers, most notably Ronald M. Kaplan, and Mental Representation is a collection of different contributions with a rather thin thread of continuity. Bresnan is the author or co-author of six of the thirteen chapters. The material is not restricted to English, and there are chapters on French, Russian, Icelandic, and Malayalam. All of the work is intended to prove applicability of the theory, rather than present the results of complete implementations.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Part III of Mental Representation is entitled \"Cognitive Processing of Grammatical Representations\"; it contains three chapters. \"A Theory of the Acquisition of Lexical Interpretive Grammars\" by Steven Pinker discusses acquisition by children. This reviewer admits to no longer being able to understand articles in this increasingly specialized sub-field of linguistics. \"A Competence-Based Theory of Syntactic Closure\" by Marilyn Ford, Bresnan, and Kaplan is, to a computationally-oriented linguist, the high point of the book. It contains a theory, supported by experimental evidence, of how real people analyze real utterances; and it ends up with a perfectly feasible plan for computerization (in terms of Kaplan's (1981) General Syntactic Processor) of lexical-functional grammars. \"Sentence Planning Units: Implications for the Speaker's Representation of Meaningful Relati0~s Underlying Sentences\" by Ford discusses production of utterances and presents some valuable experimental data.", |
| "cite_spans": [ |
| { |
| "start": 710, |
| "end": 725, |
| "text": "Kaplan's (1981)", |
| "ref_id": "BIBREF1" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "It is not possible, of course, in a review to do anything like justice to the wealth of detail in either of these books. For example, Horn systematically compares his English analysis with a parallel analysis of Polish. In fact, all of the authors wrestle with the problem of free word order and seem to achieve victories. In a chapter on \"Control and Complementation\", Bresnan tackles word order and even worse problems with some success. The language-specific chapters contain much of value that cannot be discussed here; and so on.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Lexical-functional grammar appears to be flourishing, but it has not swept away all the other schools of thought. Still other alternatives can be visualized. For example, we might question the asymmetry between phonology and semantics, and change the model to place them on a common basis. We might use phrase structure and lexicon to generate something called \"deep structure\" and two sets of descriptions, one on each side, to be solved.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Horn tries to formulate the construction of functional structures with something very like transformations rather than the arrow formalism. It is true, in an empty way, that every mapping of this kind is a transformation. But the process of solving the description wreaks havoc with conventional simplicity metrics. It is not, of course, fashionable any longer to mention simplicity metrics, but a glance at the argumentation procedures of any theoretical linguist will disclose that they are alive and well, if never acknowledged. Theoretical linguists choose, continually, between models on the basis of what mathematicians call elegance. This is unobjectionable because it is simply another version of Occam's Razor; but mathematicians live in a world where anything, notationally speaking, goes. Linguists have been, generally speaking, captive to their notational devices. It may well be that the most valuable contribution of lexical-functional grammar is to introduce the implicit function theorem into linguistics.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Book Reviews", |
| "sec_num": null |
| }, |
| { |
| "text": "Computational Linguistics, Volume 12, Numbers 2, April-June 1986", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Computational Linguistics, Volume 12, Numbers 2, April-June 1986 Book Reviews Books Received", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "The following books have been received. ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "BOOKS RECEIVED", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "A realistic transformational grammar", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Bresnan", |
| "suffix": "" |
| } |
| ], |
| "year": 1978, |
| "venue": "Linguistic Theory and Psychological Reality", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Bresnan, J. 1978 A realistic transformational grammar. In Halle, M., Bresnan, J., and Miller, G.A., Eds., Linguistic Theory and Psycholog- ical Reality. Cambridge, Massachusetts: The MIT Press.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Active chart parsing", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [ |
| "M" |
| ], |
| "last": "Kaplan", |
| "suffix": "" |
| } |
| ], |
| "year": 1981, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Kaplan, R.M. 1981 Active chart parsing. Xerox Palo Alto Research Center.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF0": { |
| "type_str": "table", |
| "html": null, |
| "content": "<table><tr><td>Department of Computer Science</td></tr><tr><td>Indiana University</td></tr><tr><td>Bloomington, IN 47405</td></tr><tr><td>THE MENTAL REPRESENTATION OF GRAMMATICAL</td></tr><tr><td>RELATIONS</td></tr><tr><td>(MIT Press series on Cambridge: The MIT Press, 1983, lii+874 pp.</td></tr><tr><td>ISBN 0-262-02158-7, $35.00</td></tr><tr><td>LEXICAL-FUNCTIONAL GRAMMAR</td></tr><tr><td>(Trends in linmdstics, Studies and monographs 21)</td></tr><tr><td>George M. Horn</td></tr><tr><td>Berlin: Mouton Publishers, 1983, 394 pp.</td></tr><tr><td>ISBN 90-279-3169-0</td></tr></table>", |
| "num": null, |
| "text": "" |
| } |
| } |
| } |
| } |
